Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
World Community Grid Forums
Category: Beta Testing Forum: Beta Test Support Forum Thread: HCC GPU Beta July 12 |
No member browsing this thread |
Thread Status: Active Total posts in this thread: 47
|
Author |
|
RetiredTech
Advanced Cruncher Canada Joined: Feb 2, 2012 Post Count: 91 Status: Offline Project Badges: |
Spot what is "delayed startup" Start Boinc after starting computer?
Bearcat do you mean the Boinc Event Log? Yes the GPU is there eventually: 12/07/2012 6:25:56 PM | | No config file found - using defaults 12/07/2012 6:25:56 PM | | Starting BOINC client version 7.0.28 for windows_x86_64 12/07/2012 6:25:56 PM | | log flags: file_xfer, sched_ops, task 12/07/2012 6:25:56 PM | | Libraries: libcurl/7.25.0 OpenSSL/1.0.1 zlib/1.2.6 12/07/2012 6:25:56 PM | | Data directory: C:\ProgramData\BOINC 12/07/2012 6:25:56 PM | | Running under account Peter 12/07/2012 6:25:56 PM | | Processor: 8 GenuineIntel Intel(R) Core(TM) i7-2600 CPU @ 3.40GHz [Family 6 Model 42 Stepping 7] 12/07/2012 6:25:56 PM | | Processor: 256.00 KB cache 12/07/2012 6:25:56 PM | | Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss htt tm pni ssse3 cx16 sse4_1 sse4_2 syscall nx lm vmx smx tm2 popcnt aes pbe 12/07/2012 6:25:56 PM | | OS: Microsoft Windows 7: Professional x64 Edition, Service Pack 1, (06.01.7601.00) 12/07/2012 6:25:56 PM | | Memory: 7.91 GB physical, 15.81 GB virtual 12/07/2012 6:25:56 PM | | Disk: 931.41 GB total, 508.72 GB free 12/07/2012 6:25:56 PM | | Local time is UTC -4 hours 12/07/2012 6:25:56 PM | | NVIDIA GPU 0: GeForce GTX 560 Ti (driver version 296.10, CUDA version 4.20, compute capability 2.1, 1024MB, 956MB available, 1382 GFLOPS peak) 12/07/2012 6:25:56 PM | | NVIDIA GPU 1: GeForce GTX 560 Ti (driver version 296.10, CUDA version 4.20, compute capability 2.1, 1024MB, 903MB available, 1382 GFLOPS peak) 12/07/2012 6:25:56 PM | | OpenCL: NVIDIA GPU 0: GeForce GTX 560 Ti (driver version 296.10, device version OpenCL 1.1 CUDA, 1024MB, 956MB available) 12/07/2012 6:25:56 PM | | OpenCL: NVIDIA GPU 1: GeForce GTX 560 Ti (driver version 296.10, device version OpenCL 1.1 CUDA, 1024MB, 903MB available) |
||
|
nanoprobe
Master Cruncher Classified Joined: Aug 29, 2008 Post Count: 2998 Status: Offline Project Badges: |
@ RetiredTech-Are you running your cards in sli mode. That is known to cause issues if you are. I would also recommend you update your drivers to 301.42 or higher. 295 and 296 drivers have issues when crunching.
----------------------------------------
In 1969 I took an oath to defend and protect the U S Constitution against all enemies, both foreign and Domestic. There was no expiration date.
|
||
|
BladeD
Ace Cruncher USA Joined: Nov 17, 2004 Post Count: 28976 Status: Offline Project Badges: |
I have had a few run for over 5 minutes with 0% progress, until I hit snooze GPU and they started running normally! Win7 with 5870s CF. More stuck at 99.415% for over 30 minutes! |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
And crash crash crash (not PC but WU)
----------------------------------------Sample: <core_client_version>7.0.25</core_client_version> <![CDATA[ <message> - exit code 293 (0x125) </message> <stderr_txt> Commandline: projects/www.worldcommunitygrid.org/wcg_beta3_img_6.47_windows_intelx86__cuda_hcc1 X0000038270735200409250017.jp2 INFO: gpu_type not found in init_data.xml. INFO: GPU device not specified in init_data.xml. Checking Commandline. ERROR: GPU device not specified in init_data.xml or on command line. Exiting. 15:45:08 (684): called boinc_finish This computer completed quite a few without incident. Like many others my rigs which had downloaded GPU WUs now have downloaded quite a bit of "extra" CPU jobs. I *think* that the reason why these crashed may be since other CPU WUs completed while these were running. The ones that returned valid did not have any other CPU WUs complete while they were running. Could the CPU thread scheduler have gotten "confused"? Merely a COMPLETE guess, but it seems weird that everything was running fine until normal CPU WUs began to complete than start new ones. Since this computer has 2 GPUs which run GPUgrid, I keep 2 threads free at all times (3820) Again, just a guess, but I think the scheduler got messed up. [Edit 1 times, last edit by Former Member at Jul 12, 2012 11:01:12 PM] |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
@RetiredTech
Not sure about BOINC versions under 7.x, but there is an option under Preference or Computing Preferences which allows you to set a delayed start (this would be when starting up BOINC from a startup). In 7.x this box is usually greyed out if GPUs are in your machines and set for use. So I'm not sure if this holds true with versions under 7. If it does not, than when a computer starts BOINC for the first time after a restart or start-up then BOINC would normally start the tasks right away. If you set the delayed start up option, this allows for the computer to "warm up" if you will, and allows BOINC to get inself going with communicating with the CPU + GPU. (not the best with pure technical speak in case you can't tell) I've seen as well as read errors about BOINC not "finding" the GPU if it starts up and begins running apps too quickly. And yes, SLI & XFire should be disabled. |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Looks like not many people had the CPU overcaching problems i told about in the beginning of the thread. So i think it's something buggy in the client version 6.10.60. To sum it up:
While the 1.00ATI+1.00CPU HCC Betas were running, Boinc kept one of 8 threads free (client set to 100% of CPUs). Whenever the estimated GPU cache went below my cache setting (0.15 days), it requested tasks for CPU+GPU, resulting in downloading a non-Beta CPU task, ignoring how big the CPU cache was already. The GPU handling only 1 task at a time can only cache up to 4 Betas, so setting the cache higher than the 4 tasks Beta cache, the client would be downloading CPU tasks endlessly. 8 GPU tasks went fine, 3 were failing (at the same second somehow). As i said before, a previously running GPU task was resumed (had been auto-suspended for a repair GPU task), and after some time it errored out and took the other 2 tasks cached up with it to errorland. Maybe an unfinished GPU task is doing something wrong to the video memory? Just a thought. Well, i keep on crunching the big CPU pile for now. |
||
|
wplachy
Senior Cruncher Joined: Sep 4, 2007 Post Count: 423 Status: Offline |
Looks like not many people had the CPU overcaching problems i told about in the beginning of the thread. So i think it's something buggy in the client version 6.10.60. To sum it up: While the 1.00ATI+1.00CPU HCC Betas were running, Boinc kept one of 8 threads free (client set to 100% of CPUs). Whenever the estimated GPU cache went below my cache setting (0.15 days), it requested tasks for CPU+GPU, resulting in downloading a non-Beta CPU task, ignoring how big the CPU cache was already. The GPU handling only 1 task at a time can only cache up to 4 Betas, so setting the cache higher than the 4 tasks Beta cache, the client would be downloading CPU tasks endlessly. 8 GPU tasks went fine, 3 were failing (at the same second somehow). As i said before, a previously running GPU task was resumed (had been auto-suspended for a repair GPU task), and after some time it errored out and took the other 2 tasks cached up with it to errorland. Maybe an unfinished GPU task is doing something wrong to the video memory? Just a thought. Well, i keep on crunching the big CPU pile for now. I also ran into to the over cache problem. Running a 0.2 day cache the 5 machines that ran Beta's ended up with a 1.6 to 9.1 day cache before I got them stopped. BOINC was requesting and receiving new CPU tasks (as you describe) and at the same time was stopping and starting High Priority CPU tasks. All in all quite a mess that's going to take a while to return to normal. BOINC client+mgr: 6.10.58 OS: Win 7 (64) and XP (32) GPU: (1) ATI and (4) NVidia
Bill P
|
||
|
Bearcat
Master Cruncher USA Joined: Jan 6, 2007 Post Count: 2803 Status: Offline Project Badges: |
Didn't see any additional wu's downloaded in 6.12.xx. Could be a client bug.
----------------------------------------
Crunching for humanity since 2007!
|
||
|
petehardy
Senior Cruncher USA Joined: May 4, 2007 Post Count: 318 Status: Offline Project Badges: |
All in all quite a mess that's going to take a while to return to normal. Normal? "Patience is a virtue", I can't wait to learn it! |
||
|
BSD
Senior Cruncher Joined: Apr 27, 2011 Post Count: 224 Status: Offline |
... When downloading the ATI driver you have to make sure the OpenCl driver is downloaded as well. I believe that this is usually found in the developer package for ATI cards. SDK? ... This was discussed already and I thought the SDK was not required as the release Catalyst already has the OpenCL driver. So do we need the developer SDK driver? I have installed the Catalyst driver and it has OpenCL: 7/11/2012 9:25:42 PM | | OpenCL: ATI GPU 0: Cedar (driver version CAL 1.4.1720 (VM), device version OpenCL 1.2 AMD-APP (923.1), 512MB, 479MB available) |
||
|
|